Graphics.DrawMeshNow + VR Single Pass Stereo

I want to draw a mesh from the viewpoint of a specific camera into a render texture using Graphics.DrawMeshNow(). I've managed to get this working when rendering to regular render texture (non-VR). However, when targeting VR render texture using Single Pass Stereo, I can't seem to figure out how to set it up properly for the mesh to be rendered into the render texture correctly.

Same goes when rendering using CommandBuffers. They work fine if using some event like AfterForwardOpaque, but when using events like AfterDepthNormalsTexture, stereo will not be set up and the render texture will be incorrectly written.

Does anyone have any experience of this?

 

I'm using custom shaders in forward rendering in built-in pipeline. There's two problems these shaders solve:

1. I'm using Graphics.DrawMeshNow() to render a custom depth texture. This depth texture is then used together with a GrabPass by an opaque object to fake transparency, while still retaining shadow support (receiving). This works fine in non-VR.

I just discovered yesterday though that it somehow seemed to work sampling a depth texture while writing into it at the same time when using forward rendering, so maybe this whole step is unnecessary. Will have to have another look at that. Regardless, it would be nice to know how to use Graphics.DrawMeshNow() with single pass stereo.

2. I'm offseting vertices in my shader, which means the internal DepthNormals shader ("Hidden/Internal-DepthNormalsTexture") will not match up with my shader. Therefore I made my own which uses command buffers. This works fine in non-VR.

That seems exactly what I'm looking for! I'll have a look at that later today.

 

Nevermind. It seems like what I did here was use an unlit (fragment) shader, that doesn't write into the depth texture since it doesn't have a shadow caster pass. As soon as I added "Diffuse" fallback, this would not work anymore. So I still need to have a way to draw a custom depth texture for VR using Graphics.DrawMeshNow().

 

I can confirm using commandBuffer.SetSinglePassStereo(SinglePassStereoMode.SideBySide) works perfectly! Object is now rendered on both sides.

So my only problem left is how to make Graphics.DrawMeshNow() do something similar to CommadBuffer.SetSinglePassStereo. Or, if not possible, how to make two invocations to Graphics.DrawMeshNow() with the correct parameters.

BTW. I had a lot of trouble setting up drawing mesh from a specific camrea, but I discovered this undocumented function which seems to work really well for this. Doesn't work with VR though!

Code (CSharp):

1. //Just an example...
2. var mat = new Material(Shader.Find("Standard"));
3. mat.SetPass(0);
4. Camera.SetupCurrent(Camera.main); // This function is undocumented, but using it will properly set up matrices for mesh to be rendered from the viewpoint of the camera.
5. Graphics.SetRenderTarget(someRenderTexture);
6. GL.Clear(true, true, Color.black);
7. Graphics.DrawMeshNow(someMesh, Matrix4x4.identity);

 

Ah, naturally I can use a command buffer for this instead. No need for Graphics.DrawMeshNow(). Just tested this and it works fine.

Would still be cool if Graphics.DrawMeshNow() was able to render in stereo, but as far as my problems are concerned, I'm all good using command buffers!